No abstract

This paper proposes a new concept for the control of voltage-source inverter (VSI)-fed induction machines. The method uses a predictive algorithm and can be split into two parts. The purpose of the first part, i.e., predictive torque control (PTC), is to predict the stator reference flux vector corresponding to the reference torque at the end of the sampling interval. The second part of the method provides accurate tracing of the stator reference flux by selecting either an active or a zero voltage vector. This approach is called immediate flux control (IFC), where two possible variants are suggested. In the first variant, a simple and fast algorithm obtains minimal stator flux error by impressing either an active or a zero voltage vector throughout the entire sampling interval. Consequently, the switching frequency becomes very low, but current and torque ripple are considerable. The second IFC variant generates the stator flux more accurately by applying an active voltage vector only throughout a calculated time slot within a sampling interval, whereas, during the remaining time of the sampling interval, a zero voltage vector is impressed. As a result, higher switching frequency arises, but it is still lower than that with space vector modulation. Both IFC variants, together with PTC, require minimal processing time and were efficiently implemented in a digital signal processor, which controlled a 3-kW induction machine drive. The obtained experimental results confirm the validity of the proposed approach.

In this paper, two novel methods for current regulation are proposed. Both methods follow the synchronized on-off principle. In the classical approach, transistors are switched depending on the sign of the current error, which in turn is sampled at equal time intervals. In the first method, the current vector at the end of the interval is predicted for two possible cases when either the active voltage vector pointing toward current error or the zero voltage vector is applied. The one producing the smaller current error at the end of the sampling interval is chosen, thus obtaining drastic reduction of the switching frequency. In the second method, the best fitting active voltage vector succeeds the zero vector during the same time interval. A simple algorithm is used to calculate the duty cycle thus gaining the smallest possible current error. The method is compared with the CRPWM. Both methods were simulated and tested on a laboratory model with passive load. In the last part of the paper, the behavior of the second method is tested for erroneously estimated load parameters. The two methods show very small degradation of performance even when a rather high parameter error is introduced.

No abstract

This paper presents a low-cost fault-tolerant system for open-phase fault in a power converter fed permanent magnet synchronous machine. The proposed fault-tolerant system is based on field orientation control with additional fault tolerance functionality. A current predictive method for open-phase fault detection is presented, together with an estimation of the threshold level for detection. The proposed method is based on the prediction of stator current for the next sampling interval. Furthermore, a new method for post-fault operation of the machine is proposed. For optimal performance of the complete drive a pre-firing angle is introduced in order to avoid the temporary generation of negative torque. This improvement increases the average generated post-fault electromagnetic torque, and consequently it reduces the mechanical stress on various machine parts. The proposed fault detection and postfault operation solutions were simulated in Matlab, and they were also tested on an experimental setup. The results show several advantages of the proposed fault-tolerant solution, like its short fault-detection time, substantial robustness against variation of machine parameters or load fluctuations, and negligible implementation costs, since no hardware modifications are needed. The fault detection algorithm does not require high computing power, and it operates well even during transients.